The rejection of sinusoidal disturbances is an active topic in the field of control. In recent years, feedforward compensation technology has been widely studied which can improve the system control accuracy. For non-minimum phase system, it is difficult to give the direct relationship between disturbance and measurable information. In addition, the frequency uncertainty and the coupling between the estimated values will add the parameter calculation burden inevitable. The coupling between the estimated values in existing adaptive methods makes it difficult to analyze the convergence performance of disturbance estimation error. At present, disturbance observer based control (DOBC) can compensate and reject multi-source uncertainty simultaneously by adjusting controller and observer parameters. Therefore, a separate observer design method is presented in this paper. Firstly, the special auxiliary filter and observer are designed to estimate the sinusoidal disturbance parameters of unknown frequency, meanwhile, the equivalent form of disturbance is reformulated. Secondly, the observer is designed to obtain the input disturbance state. With above, the nonlinear non-minimum disturbance rejection problem can be simplified to the observer design problem for linear system. The effectiveness of the method is demonstrated by Lyapunov theorem and numerical simulation.